Performance of Two-Sensors Tandem Network for Detecting Deterministic Signals in Correlated Gaussian Noise
Shailee Yagnik, R. Viswanathan, Lei Cao · 2020
In this paper, we study Bayes error performance of a two-sensor tandem network designed to detect deterministic signals in correlated Gaussian noise. Specifically, we address the question of whether the stronger-signal sensor should be the fusion center of the network for achieving the least probability of error or notƒ In the process of this query, we draw some inference parallel to the 'Good, Bad, Ugly,' signal regions formulated originally for the two-sensors one-bit-per-sensor parallel fusion network by Willett, et. al. Numerical results show that the strategy of placing stronger-signal sensor as the fusion center provides slightly better probability of error performance than the strategy of placing the stronger-signal sensor at the top of the tandem stream, for all signal points in the tandem 'Good' signal region. Significantly, optimum performance is sensitive to the decision-threshold at the top sensor, if stronger-signal sensor were placed at the top, but not when the stronger-signal sensor is at the fusion center.